An argument against the realistic interpretation of - Philsci

An argument against the realistic interpretation of the wave function
Carlo Rovelli
CPT, Aix-Marseille Université, Université de Toulon, CNRS,
Case 907, F-13288 Marseille, France.
Samy Maroun Research Center for Time, Space and the Quantum.
(Dated: August 22, 2015)
Testable predictions of quantum mechanics are invariant under time reversal. But the change of the
quantum state in time is not so, neither in the collapse nor in the no-collapse interpretations of the
theory. This fact challenges the realistic interpretation of the quantum state. On the other hand,
this fact follows easily if we interpret the quantum state as a mere calculation device, bookkeeping
past real quantum events. The same conclusion follows from the analysis of the meaning of the wave
function in the semiclassical regime.
In classical mechanics, we cannot figure out the arrow
of time from a given history.1 This is a consequence of
the invariance of classical mechanics under time reversal.
Quantum mechanics has the same property, in the following sense. If we observe the value a of an observable2 ,
the probability of observing the value b of another observable after a time t is equal to the probability of observing the value a a time t after b was observed (because
|ha|e−iHt |bi|2 = |hb|e−iHt |ai|2 ).3 More precisely: given
an ensemble of sequences of measurements outcomes, we
cannot figure out the arrow of time from them.4
However, this T-invariance is badly broken in the textbook accounts of quantum theory. To see this, consider a
system initially in a state |ai, eigenstate of an operator A
with eigenvalue a. Say at time ta , we have measured the
value a for the quantity represented by the operator A.
At a later time tb , we measure the value b for a quantity
represented by the operator B. The actual measurement
outcomes are thus a at ta and b at tb .
ta 7−→ tb
a 7−→ b.
If we reverse the history of the events, we have the time
reversed history
tb 7−→ ta
b 7−→ a.
Nothing remarkable so far. But let us now time reverse
not just the quantum events, but rather the full evolution of the state (or wave function). According to the
Unless initial data are highly non generic –for instance have low
Throughout the article I assume for simplicity all eigenvalues to
be non degenerate.
Penrose observes in [1] that T -invariance is broken by two detectors d1 , d2 separated by a beam splitter, because P1→2 = 12
while P2→1 = 1. But this is a consequence of the asymmetric
assumption that in reverse time the beam necessarily gets to d1 :
the state is fully specified at d2 but not at the d1 .
Unless the initial ensemble is highly non generic –for instance
has low entropy.
standard account, at time tb the state of the system is
projected to an eigenstate of B, say |bi, with eigenvalue
b. Assuming for simplicity the system was already in |ai
to start with, we have
ta 7−→ tb
|ai 7−→ |ai 7−→ |bi.
If we time reverse this we obtain
tb 7−→ ta
|bi 7−→ |ai 7−→ |ai
which is not what textbook quantum theory predicts!
The evolution of the quantum state predicted by standard quantum theory in association with the quantum
events (2) is rather
tb 7−→ ta
|bi 7−→ |bi 7−→ |ai,
which differs from (4) because the state of the system in
between the two measurements is |bi, rather than |ai.
Therefore: The quantum events predicted by quantum
theory (the outcomes of measurements) and their relative probabilities are T-invariant. But the evolution of
the quantum states (the wave function) in the conventional accounts of quantum mechanics is not.
A movie of a classical evolution projected backward is
still a consistent classical evolution. A movie with the sequence of the observed quantities in a series of quantum
measurements is still a consistent sequence of observations. But a movie of the changes of the wave function
evolution projected backward makes no sense.
This can be put pictorially: we can describe the βdecay of a nucleus by means of the electron wave function concentrated on the nucleus, slowly leaking out
in all directions until the electron gets detected by a
Geiger counter at some distance. The time reversed phenomenon –the Geiger apparatus emits an electron that
is then captured by the nucleus– is perfectly possible,
but it is not described by a wave function that converges
symmetrically onto the nucleus.
Notice that in all these cases there exists a consistent
quantum description of the time reversed sequence of
events (which is (5)), but at a given time between two
observed events the quantum state is different depending
on which arrow of time we assume to be the “actual” one.
But if Nature knows nothing about the “actual” arrow of
time, how can the wave function know which one is the
correct choice?
The breaking of T-invariance is due to the common
assumption that the state ends up to be projected after a measurements, and not before. While at first sight
this assumption sounds very plausible (we say that effects
come after their causes, not before), at a more careful examination this is in striking contrast with anything happening in classical mechanics, where all arrows of time
can always be traced back to special (low entropy) initial
conditions in the past. Here we appear to have a striking
violation of T-invariance in the dynamics of the state,
unrelated to any peculiar initial condition.
To put it simply: between a measurement of A and a
measurements of B what determines if the state is |ai or
What is truly striking, however, is that such a gross violation of T-invariance turns out to be unobservable! Because the actual sequence of measurements knows nothing about whether state between the measurements is |ai
or |bi. The situation is a bit like in the official Catholic
doctrine, according to which two miracles happen in the
Mass: bread is mutated in flesh, but flesh miraculously
looks like bread. Here, the quantum state knows about
the direction of time, but miraculously it keeps this info
for itself.
This breaking of T-invariance at the core of the common textbook presentation of quantum mechanics has
been emphasized by Huw Price in his remarkable book
about the arrow of time [2], where it is discussed in great
detail. I refer the reader to the book for a discussion and
replies to possible objection to the above. In the next
section I discuss what to make of it.
Heisenberg and Dirac. Example of an interpretation
based on it are Einstein’s claim that quantum mechanics is incomplete, or (maintaining the idea that
quantum theory is complete), the Relational interpretation [3–7], where the actual ontology of the
world is reduced to the sequence of the (relational)
elementary quantum events a, b ... that happen
when physical systems interact.
In this second family of interpretations, the fact that the
evolution of the state is not time reversal invariant generates no concerns: if the state is a bookkeeping device
of past event, the ambiguity in the value of the state between ta and tb reflects our choice of bookkeeping past,
or future, events.5 But if the state is taken to be real, the
fact that it behaves in a non T-invariant way, when everything we measure about the (classical and quantum)
world is T-invariant, sounds illogical.
Therefore the breaking of T invariance in the changes
of the state is a strong indication against a realistic interpretation of this state.
The fact that assigning a state to the system between
two measurements forces us to choose a direction of time
–when physics is blind to the direction of time–, strongly
indicates that the state of a system at time t is not something about the actual physics at t: it is about what we
know at that time.
There are interpretations of quantum mechanics based
on a realistic interpretation of |ψi where the collapse is
assumed not to happen. At tb , the wave function of the
system is not projected to |bi. The two main interpretations of this group are the Many Worlds interpretation
(MW) [8] and the de Broglie-Bohm hidden variable theory (dBB) [9]. These interpretations assume that during
a standard measurement what happens is only the entanglement between the system and the apparatus.
There are two main interpretations of the quantum
state |ψi of a physical system.
(i) The first is that |ψi represents the actual state of
affairs of the world. This is the realist interpretation of |ψi, which can be traced back to Schrödinger.
This point of view is common in most textbook presentations of quantum theory. It is also shared by
several no-collapse interpretations, on which I will
comment in the next section.
(ii) The second is that |ψi is just a theoretical bookkeeping devise for something else. In particular, it
can be a way of bookkeeping the real events we are
aware of. This point of view can be traced back to
|ai ⊗ |oi i 7−→ |bi ⊗ |oa i + |b0 i ⊗ |ob0 i.
In MW, the result of the measurement is assumed to be
realized in each of the branches (“worlds”). In dBB, the
ontology includes also classical-like system’s variables,
whose dynamics is guided by |ψi; these happen to follow
one or the other of the branches of the entanglement. The
“empty” branch” (|b0 i⊗|ob0 i) becomes de facto irrelevant
if there is sufficient decoherence.
How are the two sequences of events (3) and (5) described in this framework? The first gives (simplifying
We ourselves function entropically, namely exploiting past low
entropy, therefore it is not surprising that we break T invariance.
FIG. 1. Where does the breaking of T-invariance comes from,
in many worlds quantum theory?
the notation of tensor states)
ta :7−→ |aoa o0i i + |a0 oa0 o0i i
tb :7−→ |boa o0b i + |b0 oa o0 b0 i + |boa0 o0b i + |b0 oa0 o0b0 i
while the second gives
tb :7−→ |boi o0b i + |b0 oi o0b0 i
ta :7−→ |aoa o0b i + |a0 oa0 o0b i + |boa o0b0 i + |a0 oa0 o0b0 i
Again, during the intermediate between ta and tb is different in the two cases: in one case it is maximally entangled with o, in the other case it is maximally entangled
with o0 . Therefore we are in the same situation as in
the collapse interpretation: that is, the assumed evolution of |ψi depends on the time orientation even if the
empirically observable physics does not.
One way to put this observation pictorially is to look at
the cover of the book on MW [8]. See Figure 1. The cover
image is a tree, and is meant to suggest the branching of
the universal wave function. Where does this branching
comes from, if fundamental physics is invariant under
time reversal?6
A possible reply to these objections is to consider branching perspectival. The state of the universe is not branched and makes
no distinction between past and future. It is our choice that isolates a substructure of the true universal wave function, and it
is only this one that appears to display the branching and its
time arrow. In other words, the true state of the system we observe in the laboratory is highly entangled in the past as well
as the future, but we ignore the aspects of this state which are
not relevant to our interest in predicting future evolution. The
price to pay for taking this view is that the wave function we
use in a concrete utilizations of the theory is not the real wave
function anymore. It is a theoretical booking devise about some
limited information we happen to have about the actual state of
the world, because of some past interactions. But if so, then the
“real” state of WM and dBB is reduced to a hypothetical metaphysical entity about which we know very little; the quantity we
call “state” when we use quantum theory, instead, turns out to
be precisely a mere bookkeeping device for our time-oriented limited information! Along this line, MW and dBB start sounding
like Lao Tzu [10]: “The Wave Function that can be told of, is
not the True Wave Function”...
In [2], Huw Price chooses a realistic interpretation of
the wave function, but explores a different solution to the
problem of the breaking of T-invariance. Price makes the
hypothesis that there is a (hidden variable) state, well defined between any two measurements, which is not determined solely by the past measurement: it is determined
also by the future measurement. This solution implies a
form of retrocausality: what happens at time t can be
determined also by later events.
This is different from classical theory, where what happens is entirely determined locally either by the past or
by the future. In Price proposal we have to assume that
to determine locally what happens we have to know both
the past and the future. Retrocausality is realized in
classical mechanics as much as forward causality: since
classical mechanics is T-invariant, the state of physical
systems is equally determined by what happens before
or by what happens later. An arrow of time arises in
classical mechanics only insofar uncommon past conditions are realized. In a relativistic theory, in particular,
the state of the world on a spacial region R determines
uniquely everything in the future domain of dependence
of R, as well as everything in the past domain of dependence of R. Price suggestion, instead, is to build a
hidden variable theory where the state between two measurements does not depend on one of the two, but rather
on both.
The problem with this is that the state of the world at
a given time, in a given region, is not sufficient to determine the state of the world in future, or past, domains of
dependence. This is a violent form of violation of locality, which, again, appears to contradict the empirical fact
that, according to quantum theory, the probability distribution of all measurements performed in the domains of
dependence of a spatial region R are uniquely determined
by (sufficient) measurements in the region R. Once again,
one adds to the postulated real world a strange phenomenon, never observed in the universe, which however
is then absent in the observed world! Again like the real,
but invisible, “Flesh” of the note.
If instead we renounce the idea that the quantum state
represents the actual real state of the world at all times,
and we interpret it simply as a way of coding what we
happen to know about the (past or future) interactions
of the system, then reality, much simplified, and formed
simply by actual quantum events, is simply T-invariant,
and subject to an entirely local dynamics. This is for
instance the ontology of the Relational interpretation.
The observed dynamics of the world is time-reversal invariant: a given un-oriented sequence of quantum events
does not determine a time arrow. Charging the wave
function, (more in general, the quantum state) with a
realistic ontological interpretation, leads to a picture of
the world where this invariance is broken. It is not broken in the sense that the full theory breaks T -reversal
invariance (it does not), but in the sense that the wave
function we associate to observed events depends on a
choice of orientation of time.
Instead, restricting the ontology to the instantaneous
(relational [3]) quantum events happening at interactions, saves us from such baroque constructions.
In general, we can get clarity about the meaning of a
quantity in an advanced theory by considering the role
of this quantity in a regime where the theory reduces to
physics we understand. This can be done for the wave
function by looking at the semiclassical limit of quantum
mechanics. In this limit, the wave function is the (exponent of the) Hamilton function. The Hamilton function
is a quantity that nobody would dream of charging with
an ontological meaning. It is a technical device for computing predictions about real events. The same is true
for the quantum state at some time. It is not a representation of reality at that time: it is a theoretical tool
for computing probabilities about possible real quantum
These quantum events provide a reasonable ontology
for quantum theory. Their characteristic aspect is to
be discrete. Discreteness has been the first manifesta-
[1] R. Penrose, The Road to Reality : A Complete Guide to
the Laws of the Universe. Knopf, 2005.
[2] H. Price, Time’s Arrow. Oxford University Press, 1996.
[3] C. Rovelli, “Relational Quantum Mechanics,” Int. J.
Theor. Phys. 35 (1996) 1637, arXiv:9609002
[4] C. Rovelli, “*Incerto tempore, incertisque loci*: Can we
compute the exact time at which a quantum
measurement happens?,” Found. Phys. 28 (1998)
1031–1043, arXiv:9802020 [quant-ph].
[5] M. Smerlak and C. Rovelli, “Relational EPR,” Found.
Phys. 37 (2007) 427–445, arXiv:0604064 [quant-ph].
tion of quantum theory (quanta of light, discrete spectra, discrete angular momentum, particles in quantum
fields...) and discreteness remains the core physical aspect of quantum theory: the number of distinguishable
states in a finite phase space volume is finite –determined
by the volume in ~ units. Forgetting discreteness and focusing only on the continuity of ψ, as a number of nowadays popular interpretations of quantum theory do, is
misleading. Quantum mechanics, as the name indicates,
is about “quantum” –that is “discrete”– events.
God did not over-fill the world with moving waves on
infinite dimensional configuration spaces. She didn’t even
draw it with heavy continuous lines. She just dotted it
with sparse quantum events.
FIG. 2. She just dotted it with sparse quantum events.
[6] B. C. van Fraassen, “Rovelli’s World,”Foundations of
Physics 40 (Apr., 2010) 390–417.
[7] M. Dorato, “Rovelli’ s relational quantum mechanics,
monism and quantum becoming,” arXiv:1309.0132.
[8] S. Saunders, J. Barrett, A. Kent, and D. Wallace, Many
Worlds?: Everett, quantum theory, and realism. Oxford
University Press, 2010.
[9] S. Goldstein, “Bohmian Mechanics,”.
[10] Lao Tzu, Tai Te Ching. Hackett, 1993.